Method configure

src/cpu/kernels/CpuTransposeKernel.cpp:734–765

} // namespace

void CpuTransposeKernel::configure(const ITensorInfo *src, ITensorInfo *dst)
{
    ARM_COMPUTE_TRACE_EVENT(ARM_COMPUTE_PROF_CAT_CPU, ARM_COMPUTE_PROF_LVL_CPU, "CpuTransposeKernel::configure");
    ARM_COMPUTE_ERROR_ON_NULLPTR(src, dst);

    // Destination auto inizialitation if not yet initialized
    const TensorShape dst_shape = misc::shape_calculator::compute_transposed_shape(*src);
    auto_init_if_empty(*dst, src->clone()->set_tensor_shape(dst_shape));

    // Explicitly set the tensor shape to preserve dimensions
    dst->set_tensor_shape(dst_shape);

    // Perform validation step
    ARM_COMPUTE_ERROR_THROW_ON(validate(src, dst));

    // Note: This kernel performs 16 elements per iteration.
    // However, since we use a left-over for loop on both dimensions (X and Y), we cannot have any read or write out of memory
    // For this reason num_elems_processed_per_iteration_x is set to 1
    const unsigned int num_elems_processed_per_iteration_x = 1;
    const unsigned int num_elems_processed_per_iteration_y = num_elems_processed(src->element_size());

    // Configure kernel window
    Window win =
        calculate_max_window(*src, Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));

    // The CpuTranspose doesn't need padding so update_window_and_padding() can be skipped
    Coordinates coord;
    coord.set_num_dimensions(dst->num_dimensions());
    dst->set_valid_region(ValidRegion(coord, dst->tensor_shape()));

    ICpuKernel::configure(win);
}

Status CpuTransposeKernel::validate(const ITensorInfo *src, const ITensorInfo *dst)
{